Field Notes
Essays from inside a 120-person digital agency on Jira — written by Ernest. Project margin, estimating, the architecture of profitability, and the small decisions that compound. No content marketing.
A first-person decision guide on when to hire a finance director at a digital agency: the ladder from bookkeeper to management accountant to fractional FD to full-time CFO, mapped to size and the cadence of decisions you actually need.
We had been carrying three retainers for years that the spreadsheet pack reported as marginally profitable. Once the live margin layer was in, every one of them was negative. We cancelled all three inside a quarter. The shape of the conversations, and what an agency owner discovers about themselves when the numbers stop hiding.
An anonymised post-mortem on a project that looked profitable on the invoice and bled quietly underneath. The exercise that turned the 50-person agency I was CTO of into a 120-person agency that knows where its margin actually goes.
A six-person team idle for a fortnight costs about £17,000, and that cost appears on no report — the bench dissolves into overhead and quietly degrades the margin of every healthy project. How to price idle capacity, why discounted work can cost more than the bench it fills, and what the bench is actually telling you.
A £450/day contractor and a £340/day employee are not £110 apart — they are two different cost shapes, and agencies routinely misprice both. When the contractor is genuinely cheaper, when they quietly eat the margin, and why contractor cost so often never reaches the project report at all.
Overhead is the cost line most agencies never allocate to projects at all — which is how a book of 50% projects produces an 8% year. The pro-rata method, the minimum-revenue floor that stops slow months distorting it, and a worked example of one project in a strong month and a weak one.
Finance needs to read Jira — worklogs, estimates, project structure — but a CFO with a Jira admin account is the wrong answer to the right question. How agencies give finance a real view of delivery data: scoped accounts, dashboards, and a financial layer where every role reads the same numbers in the unit that fits their job.
The delivery experience of Jira Cloud and Jira Data Center is nearly identical. The finance experience isn't. Authentication, worklog APIs, data residency and plugin economics all differ — and they decide which financial tooling your agency can actually run. A practical comparison for agency CTOs and finance leads.
Jira projects are organised for delivery — one key per client, per team, per product. Your P&L is organised by contract. The two almost never line up, which is why agencies with perfect worklogs still can't answer what a specific engagement earned. Here is the mapping layer that fixes it.
Every margin number is downstream of the worklogs, and worklogs go wrong for predictable reasons — most of them incentives, not laziness. Six steps to time data you can trust: what to ask for, what to stop rewarding, and why surveillance produces worse numbers, not better ones.
15 minutes on your real Jira, our screen.
Book a demoMost agencies discover a project was unprofitable at the monthly close — weeks after the costs landed, often after the team has moved on. It isn't a speed problem with your accountants; it's a cadence problem with the number itself. What lateness costs, and what changes when margin is continuous.
Scope creep runs at 5–12% of contract value on friendly projects — which sounds survivable until you notice it comes straight out of margin, not revenue. On a 25%-margin project, a 10% creep takes nearly half the profit. Where it hides, a worked example, and how to count it.
Salary divided by 2,080 is not what an employee costs you per hour — the real figure is often 60–80% higher once employer taxes, pension, per-head costs and paid non-working days are counted. A step-by-step calculation with a worked example for a £55,000 developer.
On a real-margin basis — labour at true cost, overhead allocated, unbilled scope counted — a healthy agency project lands around 20–30%. Most agencies quoting 50–60% are quoting labour margin, a different number. The honest benchmark ranges, and why the popular ones mislead.
Most agencies that lose money on projects don't know it — the signals are operational, not financial, and they show up months before the accounts do. Seven signs to check against your own agency, and the data question that confirms or clears each one.